Multi-robot Markov random fields

Abstract

We propose Markov random fields (MRFs) as a probabilistic mathematical model for unifying approaches to multi-robot coordination or, more specifically, distributed action selection. The MRF model is well-suited to domains in which the joint probability over latent (action) and observed (perceived) variables can be factored into pairwise interactions between these variables. Specifically, these interactions occur through functions that evaluate between an observed and latent variable and between a pair of latent variables. For multi-robot coordination, we cast local evidence functions as the computation for an individual robot's action selection from its local observations and compatibility as the dependence in action selection between a pair of robots. We describe how existing methods for multi-robot coordination (or at least a non-exhaustive subset) fit within an MRF-based model and how they conceptually unify. Further, we offer belief propagation on a multi-robot MRF as a novel approach to distributed robot action selection.